TrustedFace - Technologie pro prevenci bezpečnostních útoků skrze DeepFakes a FaceMorph

The project's mission is to develop methods for detecting fake and manipulated images of human faces that could be used to compromise system security. The first objective is to survey available attack methods and compile a dataset of fake images and video sequences suitable for further development as well as for informing both the professional community and the general public. The second objective is to develop methods for detecting deepfake images and videos capable of facilitating identity spoofing. Such spoofing can target automated facial recognition systems (e.g., building access control or biometric ticketing systems for sporting events) or individuals (e.g., a fraudulent video call with a legal representative intended to elicit sensitive information or issue unauthorized instructions). The third objective is to develop methods for detecting morphed images, which attackers might use to falsify their identity when obtaining authentic, biometrically secured documents. The fourth objective is to create a pre-production solution based on the developed methods and to build a demonstrator; this will serve to validate the solution's capabilities and provide material for marketing communications aimed at the applicant's clients and the general public.